Jovian revolutionizes variant detection in Next-Generation Sequencing data using generative AI, offering state-of-the-art accuracy.

The current challenge is to reduce the false positive rate in variant detection crucial for downstream analysis, overcome the complexities and limitations of statistical models, and address the difficulties in detecting small variants in the vast data produced by next-generation sequencing.

Jovian represents a breakthrough in genomic technology, specifically engineered to detect small genetic variants within the massive datasets generated by Next-Generation Sequencing (NGS). This advanced system departs from conventional methods by adopting a generative AI framework that employs a transformer-based deep learning model. This model is adept at accurately predicting true sequences from NGS reads, thereby pinpointing variants with remarkable precision. Jovian's capabilities are pivotal for genomic research and diagnostics, particularly for uncovering genetic disorders and predispositions. It also plays a crucial role in pharmaceutical research, aiding in drug discovery and the tailoring of personalized medicine. Moreover, Jovian enhances bioinformatics tools and services, significantly improving upon existing NGS data analysis platforms. In the realm of clinical genomics, Jovian stands out by bolstering the accuracy of genetic testing, making it an invaluable asset in the continuous pursuit of medical innovation.

Clinical Validation

• Utilizes generative AI for accurate variant prediction.
• Employs a single, transformer-based deep learning model, simplifying the variant detection process.
• Achieves state-of-the-art performance in accurately detecting small variants.
• Scalable model size and training data quality and quantity.
• Offers an innovative approach compared to conventional, statistically-based methods.

Brendan O’Fallon, Ashini Bolia, Jacob Durtschi, Luobin Yang, Eric Frederickson, Katherine Noble, Joshua Coleman, Hunter Best, Jovian enables direct inference of germline haplotypes from short reads via sequence-to-sequence modeling, bioRxiv 2022.09.12.506413; doi: https://doi.org/10.1101/2022.09.12.506413

Aaron Duffy
(801) 585-1377
aaron.duffy@utah.edu